I had always some loose thought in my mind which I’m eager but not able to systemically answer, which is what problem could/couldn’t be well solved by model machine learning/deep learning etc methodologies. In particular, I want to claim (loosely) that even if one can find the most powerful machine learning, it won’t be useful in predicting the financial or economics behavior in the world because it’s essentially ruled by game theory/mean field game theory, i.e., the outcome of the result really depends on the other players in the game so that no model could ‘predict’ it.
Admittedly, one can always give some frameworks to claim the effectiveness of ML in financial world, however, I’m eager to see very simple naive models to formalize this problem (either support or against my claim) in an elegant way.
VictorX 